Viral Recipe Discovery Agent

Multi-lingual recipe discovery & sentiment analysis

Media & FoodViral Recipe Discovery Agent
IndustryMedia & Food
Impact
  • Recipe output scaled from 40 to 1,000+ per month
  • App churn cut from 28% to 12%
  • Curation time down from 10+ hours to 2 hours per week

Overview

A German recipe media company with 500K+ newsletter subscribers needed a constant stream of proven air fryer and Thermomix recipes from social media, structured, deduplicated, and published in two languages.

The challenge

Manual curation could not feed the audience.

  1. Curation took 10+ hours a week and still produced only about 40 recipes a month.
  2. No way to tell which social recipes actually resonated: engagement and audience sentiment went unmeasured.
  3. Every recipe needed structuring, nutrition data, and dual-language output in English and German, all by hand.

The approach

  1. Monitors 200+ creator profiles and 50 hashtags across TikTok and Instagram for new recipe content daily.
  2. Captures each video with its transcript and engagement data, then scores virality on weighted shares, comments, likes, and saves.
  3. Runs sentiment analysis on audience comments to grade reception and usefulness before anything is kept.
  4. Generates the structured recipe in English and German with SEO metadata, straight into a database that carries the engagement metrics.
  5. Duplicate detection stops the same recipe from being processed twice.

The unlocked content volume enabled personalized in-app recommendations, which is what moved churn from 28% to 12%.

The results

  1. Recipe output scaled from 40 to 1,000+ per month
  2. App churn reduced from 28% to 12% on the back of personalized recommendations
  3. Curation time cut from 10+ hours to 2 hours per week
  4. Every recipe delivered in English and German with SEO metadata, no translator involved
  5. Structured recipe database with engagement metrics attached to every entry
  6. Duplicate detection prevents wasted processing
  7. Handed over with an operator guide and technical deep dive, so the team runs it in-house

System snapshots

01Recipe Content Database
Master database holding every discovered recipe with Google Doc URL, top-performer flag, cuisine type, source video, status pills, and output-generation state per row.
Recipe Content Database
02Creator & Profile Watchlist
Per-creator tracking across TikTok and Instagram, segmented by channel and active state, so the agent only ingests fresh content from accounts that are still publishing.
Creator & Profile Watchlist